Hi5: Robust 2D Hand Pose Estimation for Real-World Applications Using Synthetic Data Alone

Dataset Download:

Link: https://www.dropbox.com/scl/fo/cs5bdynig2a8jazpomul7/AFJ_f9mKHxlsrR60eGn0SRU?rlkey=tw1y19spyxtsca94pz6on2z1j&st=uduvogbm&dl=0

Unity Game Engine Project

Same dropbox link above, Hi5_Unity_Project.zip
For game engine project setup, please follow unity-setup-guide.md

Data Augmentation

The data_processing/ directory contains code for processing, and augmenting the synthetic data.

Training Code

For Training and Evaluation, please use ViTPose repository: https://github.com/ViTAE-Transformer/ViTPose
Please follow the training and evaluation instructions from the repository.
To reproduce our results, add the content in the config files to ViTPose/configs/hand/2d_kpt_sview_rgb_img/topdown_heatmap/
Exact command line arguments used in this study can be seen in terminal commands.md file.
Trained checkpoints in Dropbox

Other datasets

OneHand10K: https://www.yangangwang.com/papers/WANG-MCC-2018-10.html
11k hands: https://sites.google.com/view/11khands